Forecast Sensitivity to satellite observations in global and East-Asia
Introduction
summer
COMS
winter
Evaluation methods of the observation impacts on NWP system
WV
▶ Adjoint-based FSO (Forecast Sensitivity to Observations) method
0300-top_IR
IR
- Reduces the computational costs of the observation impact calculations
~74%
0300-top_IR
~42%
~52%
Experiment Design
HIRS_5
AMSUA_N18
AMSUA_MetB
HIRS_6
HIRS_6
AMSUA_N19
HIRS_6
AMSUA_MetA
HIRS_7
HIRS_7
AMSUA_MetA
HIRS_7
AMSUA_N19
HIRS_11
HIRS_11
AMSUA_N15
HIRS_11
AMSUA_N15
HIRS_12
HIRS_12
AMSUB_N18
HIRS_15
HIRS_15
AMSUB_N19
AMSUB_3
AMSUB_3
AMSUB_MetB
AMSUB_4
AMSUB_4
AMSUB_5
AMSUB_MetA
AMSUB_5
AMSUA_1
AMSUA_1
AMSUB_N19
AMSUA_2
AMSUA_2
AMSUA_4
AMSUA_4
AMSUA_5
AMSUA_5
AMSUA_5
AMSUA_6
AMSUA_6
AMSUA_6
AMSUA_7
AMSUA_7
AMSUA_8
AMSUA_8
AMSUA_9
AMSUA_9
AMSUA_10
AMSUA_10
AMSUA_11
AMSUA_11
AMSUA_12
AMSUA_12
AMSUA_13
AMSUA_13
AMSUA_14
AMSUA_14
AMSUB_N18
• December 2013 ~ February 2014 for winter
-6.E-06
-4.E-06
-2.E-06
Observation data used in KMA
local tovs
Surface : Synop, Ship, Buoy, METAR, BOGUS
Sonde : Radiosonde, PILOT, Wind profiler, DROPSONDE
Aircraft : AMDAR, AIREP
Scatwind : ASCAT (MetOp2-A, MetOp1-B)
AMV (Atmospheric Motion Vector) : GOES, AVHRR, MODIS, COMS (KMA), MTSAT (JMA), Meteosat7,
Meteosat10
ATOVS : NOAA15, 18, 19, MetOp2-A, MetOp1-B
IASI : MetOp2-A, MetOp1-B
※ KMA has been used the MetOp1-B satellite
AIRS : AQUA
CSR : COMS (KMA)
since November, 2013
GPSRO : GRAS, GRACE, COSMIC
global tovs
-2.E-05
Impact for
satellite
-9.0E-06
-6.0E-06
Impact for
satellite channel
-1.2E-04
-9.0E-05
-3.0E-06
-3.0E-05
Meteosat10
AMSU-A
1.E-06
WV
NOAA15
MTSAT
AQUA
NOAA18
-2.0E-05
HIRS and CLR
-1.5E-05
-9.9E-06
VIS
GPSRO
AMSU-B/MHS
ASCAT
AIRS
AMSU-A
WV
0.0E+00
-8.0E-05
Similar pattern
AIRS
-4.9E-06
-6.0E-05
-4.0E-05
AMV
MTSAT
~ 68%
AMSUA_12
AMSUA_13
AMSUA_13
AMSUA_14
AMSUA_14
-3.E-06
0.E+00
3.E-06
0.0
0.5
1.0
1.5
25%
COMS
Window
(648 ~ 759 cm-1)
(806 ~ 962 cm-1)
~ 51%
VIS
MetOp
Negative
279
277
279
277
275
273
271
263
259
221
202
200
195
186
184
179
176
170
165
163
145
140
137
135
132
130
127
125
123
121
119
117
115
113
111
109
107
105
103
101
99
97
95
93
91
88
86
84
82
80
78
76
74
72
70
68
66
64
62
60
58
56
54
51
47
42
37
32
27
Positive
(1014.5 ~ 1062.5 cm-1)
Sea
Window
WV band
(806 ~ 962 cm-1)
(1212.5 ~ 2034.25 cm-1)
273
275
277
279
275
277
279
271
273
263
259
221
202
200
195
186
184
179
176
170
165
163
145
140
137
135
132
130
127
125
123
121
119
117
115
113
111
109
107
105
103
101
99
97
95
93
91
88
86
84
82
80
78
76
74
72
70
68
66
64
62
60
58
56
54
51
47
42
Positive
HIRS, CLR and AIRS
GPSRO
IASI
AMSU-B/MHS
ASCAT
WV
- In window and O3 band, IASI degrades the forecast skill
- There are neutral impacts on WV band
- Distinct differences between Land and Sea appear at lower stratosphere (some land channels give negative impacts)
VIS
For winter
IR
- There are generally positive impacts on CO2 band for both Sea and Land (similar to summer)
AMSU-A
- Positive impacts are reduced/increased at middle-low troposphere on Sea/Land compared to summer
1.0E-07
- There are neutral impacts on Window, O3 and WV band
4
Conclusions and Future plan
 Impacts of MetOp satellite and AMSU-A channel among satellite observations are large for both
 Impacts of MTSAT and COMS satellite (AMV) are large for East-Asia area due to the their
mainly distributions
IR
VIS
VIS
WV
~66%
-1.5E-05 -1.0E-05 -5.0E-06 0.0E+00
♣ MTSAT > COMS > Meteosat7 > GOES (+ Polar wind) > Meteosat10
♣ IR > WV > Visible channel for total impact
-1.0E-05
-5.0E-06
0.0E+00
 For ATOVS, mean impacts of NOAA18 are largest among satellites and MetOp1-B (started to be used
November 2013) gives positive impacts
 For IASI, there are generally positive impacts on CO2 band for both Sea and Land
- need to further analysis about channels which degrades forecast skill
◆ Future Plan
 Evaluation of each AIRS channel
 Development of local-area FSO to evaluate the observations for local-area
271
263
259
221
202
200
195
186
184
nearly neutral
179
176
170
165
163
145
140
132
130
127
125
123
121
119
117
115
111
109
107
105
103
101
99
97
95
93
91
88
86
84
82
80
78
76
74
72
70
68
66
64
62
60
58
Positive
137
nearly neutral
135
Land
- There are generally positive impacts on CO2 band for both Sea and Land (Especially middle of troposphere)
IR
- For summer/winter, impact per observation is largest in WV/IR channel
275
Negative
- There are negative impacts at lower stratosphere
~ 35%
Meteosat7
273
Negative
For summer
NOAA18
WV
MTSAT
~51%
271
263
259
221
202
200
195
186
184
179
176
170
165
163
145
140
137
135
132
127
125
130
Negative
123
121
119
117
115
113
111
109
107
105
103
101
99
97
95
93
91
88
86
84
82
80
78
76
74
72
70
68
66
64
62
60
58
56
54
51
47
42
37
32
27
(1212.5 ~ 2034.25 cm-1)
-9.E-06
COMS
IR
2.5
WV band
Meteosat7
GOES
WV
IR
2.0
◆ Need to additional analysis and discussions
- Impact per observation
: local ATOVS (RARS) > global ATOVS
- O-B increase ↔ negative impact increase
CO2 band
-6.E-06
NOAA19
winter
VIS
~40 %
-6.E-06
AMSUA_12
-3.E-06
NOAA15
Meteosat10~0%
~19%
-9.E-06
AMSUA_11
AMSUA_11
global and East-Asia area
GOES
WV
2.5
0.0E+00
CO2 band
56
-1.E-21
-2.E-06
1.0E-07
-2.0E-05
summer
COMS
2.0
-5.0E-06
0.E+00
Especially for East-Asia area
Meteosat7
AMSUA_10
O3
37
IR
Results –East Asia
Meteosat10~0%
1.5
-1.0E-05
(648 ~ 759 cm-1)
Impacts of MetOp satellite and AMSU-A channel are large Impacts of MTSAT and COMS satellite (AMV) are large
Impacts per observation are large in GPS-RO and IASI
- Due to the their mainly distribution
3
AMSUA_9
-1.4E-19
GPSRO
IASI
IR
AMSU-A
VIS
WV
ASCAT
AMSU-B/MHS
AIRS
COMS CLR
HIRS
IR
IASI
1.0
-1.5E-05
Positive
32%
NOAA18
Meteosat7
NOAA19
NOAA15
GOES
AQUA
Meteosat10
NOAA19
MetOp
AMSUA_8
AMSUA_9
winter
54
-2.0E-05
6
2
IASI
Meteosat10, AQUA and
GOES
MTSAT
COMS
MTSAT
MetOp
0.0E+00
GPSRO
IASI
AMSU-A
WV
AIRS
ASCAT
IR
VIS
AMSU-B/MHS
HIRS
COMS CLR
-6.0E-05
GOES
AMSUA_7
AMSUA_7
-2.E-06
VIS
51
-4.0E-05
COMS
Meteosat7
GOES
MetOp
NOAA18
NOAA15
AQUA
MTSAT
Meteosat7
NOAA19
Meteosat10
COMS
AMSUA_6
ASCAT
GPSRO
Impact for East Asia
Impact per observation
Total Impact
Impact for Global
Impact per observation
Total Impact
AMSUA_5
-1.E-06
-1.E-06
3.E-06
winter
-1.2E-05
HIRS and CLR
47
-6.0E-05
0.5
Positive
42
-8.0E-05
AMSUA_4
2.E-06 AMSUA_4
0.E+00
HIRS
AMSU-B/MHS
COMS CLR
AIRS
3.0E-05
0.E+00
AMSUB_5
(1014.5 ~ 1062.5 cm-1)
MetOp
32
20 %
IR
-2.0E-05
AMSUB_5
O3
-2.E-06
1.E-06
37
-7.0E-05
HIRS_MetA
AMSUB_4
nearly neutral
32
-1.2E-04
-2.0E-05
Land
Meteosat7
GPSRO
IASI
WV
IR
AIRS
AMSU-A
VIS
ASCAT
WV
AMSU-A
NOAA18
COMS
AMSU-B/MHS
AMSU-B/MHS
GPSRO VIS
ASCAT
satellite channel
AMSUB_4
AMSUB_3
-1.E-06
2.E-06
1.0E-07
HIRS and CLR
Impact for
AQUA
27
-4.9E-06
MTSAT
27
-9.9E-06
30%
6
NOAA18 NOAA15
IASI
HIRS_15
AMSUA_10
0.E+00
NOAA15
6
AQUA
GPSRO
IASI
AMSU-A
AIRS
IR
WV
ASCAT
VIS
AMSU-B/MHS
HIRS
COMS CLR
HIRS_MetB
Negative
6
Meteosat7
MTSAT
-1.5E-05
AMSUB_3
global tovs
3.E-06 0.0
1.E-06
HIRS_12
AMSUA_8
NOAA19
MTSAT
COMS
MetOp
Meteosat7
NOAA18
AQUA
NOAA15
NOAA19
GOES
Meteosat10
NOAA19
1.7E-21
-1.E-06
1.E-06
2
-3.0E-06
-3.E-06
Sea
2
-6.0E-06
2.E-06
2
-9.0E-06
0.E+00
HIRS_11
HIRS_15
summer
Meteosat10 and GOES
Meteosat10
MetOp
-5.E-06
HIRS_7
IASI
COMS
GOES
-1.E-05
HIRS_6
HIRS_12
local tovs
♣ AMSU-A > HIRS > AUSM-B/MHS
♣ NOAA18 > Metop > NOAA19 > NOAA15
♣ MetOp1-B gives large positive impacts
Impact for East Asia
Impact per observation
Total Impact
Impact for Global
Impact per observation
Total Impact
-2.E-05
-5.E-06
Results – Global VS East Asia
-1.2E-05
0.E+00
HIRS_5
HIRS_5
AMSUB_MetA
-1.E-05 -8.E-06 -6.E-06 -4.E-06 -2.E-06
HIRS_4
HIRS_4
HIRS_5
HIRS_MetA
-8.E-06
O-B
AMSUA_N18
HIRS_N19
• June ~ August 2013 for summer
Impact per observation
O-B
HIRS_4
Data period
satellite
winter
HIRS_4
Reduction in error variance dry energy norm, surface to ~150 hPa, for 24-hour forecast
Impact for
-4.0E-05 -3.0E-05 -2.0E-05 -1.0E-05 0.0E+00
summer
Observation Impact measure
GOES
MetOp
NOAA18
NOAA15
AQUA
MTSAT
Meteosat7
NOAA19
Meteosat10
COMS
below-700_Visible
Impact per observation
summer
below-700_IR
below-700_IR
ATOVS
▶ Detail analysis the results of East Asia FSO
3
~67%
below-700_Visible
0700-500_IR
~19%
♣IR channel gives largest impacts for both COMS and MTSAT
- Portion of IR impact decreased in winter for both COMS and MTSAT
♣ Negative impacts are shown in East China area
♣ For summer, MTSAT gives larger positive impacts at mid-low (700~500 hPa) levels than COMS
♣ For winter, impacts per observation of COMS are larger than MTSAT especially at low~mid. levels (~500 hPa)
▶ Comparing the impact of Global and East Asia area
• 6 hourly (00Z,06Z,12Z,18Z)
VIS
IR
-4.0E-05 -3.0E-05 -2.0E-05 -1.0E-05 0.0E+00
• Especially focused on…
- Resolution: UM N512, 4D-Var N216
negative
0500-300_IR
0700-500_IR
VIS
• Evaluation of satellite impacts by using KMA FSO system
- Version 7.9 PS28 – operational from 27 June 2013
0300-top_WV
WV
0500-300_IR
large
positive
WV
IR
Object
•
•
•
•
•
0300-top_IR
0300-top_WV
Numerical forecast error reduction/increase due to the observation use
•
•
•
•
•
-4.0E-05 -3.0E-05 -2.0E-05 -1.0E-05 0.0E+00
-4.0E-05 -3.0E-05 -2.0E-05 -1.0E-05 0.0E+00
▶ What is FSO?
• Met Office global Unified Model (UM) with 4D-Var
0700-500_IR
below-700_Visible
below-700_Visible
MTSAT
Use an adjoint technique in VAR system
Trace it back to the observations used in the analysis
Forecast error : Energy Norm (J/kg)
Negative value means error reduction → good impact
Impact of each observation type (sorted by channel, location, variable, vertical level, etc)
written in a single ascii file
negative
below-700_IR
79%
below-700_IR
observations grouped by instrument type, variable, region, vertical level, etc
NWP system
IR~
0700-500_IR
- Observation impact can be efficiently estimated for a set of observations or any subset of
2
negative
0500-300_IR
large
positive
~17%
0500-300_IR
~16%
0300-top_WV
VIS
0300-top_WV
VIS
- Not necessary to add or remove observations from the assimilation to estimate their impacts
-
0300-top_IR
WV
113
1
Eun-jung Kim, Sangwon Joo, Yoonjae Kim and Dong-Joon Kim
Numerical Data Preprocessing Team
Korea Meteorological Administration